RNA-protein interactions are at the heart of many cellular processes, and an important goal is to understand the molecular details that govern their specificity. Much previous work has focused on small model systems in which arginine-rich peptides derived from viral regulatory proteins (including HIV- 1 Tat and Rev and BIV Tat) recognize their RNA hairpin targets in the absence of a surrounding protein framework. The results show that arginine-rich peptides can adopt very different secondary structures and often require the framework of the RNA for folding. To better understand how amino acid-RNA interactions contribute to RNA-binding specificity in different contexts, from small peptide-RNA complexes to large multiprotein complexes, we now propose to: 1) use combinatorial strategies to more precisely define molecular details of amino acid-RNA interactions in small model systems and identify novel interactions, and 2) define how RNA-binding specificity is established and regulated in the context of multiprotein complexes in vivo, using recognition of branchpoint sequences at 3' splice sites as a modeI system. Specific Aim 1 wifi utilize the arginine-rich and zinc fmger motifs to generate combinatorial libraries and will use a bacterial antitermination system to identify peptides that bind to the RRE RNA and mutant sites; This Aim also wifi experimentally test amino acid-base pair interactions predicted by computer, and in particular an arginine-GU interaction involving three hydrogen bonds and an altered specificity variant of the Rev-RRE interaction. Specific Aim 2 will examine how specificity is determined for mamnialian branchpoint sequences at 3' splice sites, in the context of a multiprotein complex involving SF1/mBBP. bound at the branchpoint and U2AF bound at an adjacent polypyrimidine tract and AG dincucleotide. Amino acids in the KB domain of SF1/mBBP important for branchpoint recognition will be identified using a Tat-hybrid system in which proteins fused to the activation domain of HIV-1 Tat are delivered to RNA sites in HIV reporter plasmids, and attempts will be made to generate change-of-specificity mutants. The Tat-hybrid system also will be used to identify other proteins from cDNA libraries that may alter branchpoint sequence recognition or that may interact with alternatively spliced forms of SF1/mBBP. The proposed. Aims are expected to add to the basic understanding of how amino acids interact specificially with RNAs and how different protein contexts may alter these recognition Properties, Dotentiallv influencing 3' splice sitea selection in the SF1/mBBP model system.